This research tests a theoretical model proposing that low competence is a basic factor in early-onset adolescent substance use. The model proposes that the combination of low academic and social competence is a risk factor that is mediated through increased levels of negative affect and affiliation with drug-using peers. It is proposed that low competence is related to background factors of life stress and temperament, but these may be moderated by adaptive coping and high family support. These hypotheses are tested in a 3-year longitudinal study with a sample of N = 2,000 students, first surveyed at the beginning of 7th grade and followed over the period through end of 9th grade. A self-report questionnaire, administered in classrooms once per year, provides data on predictor variables; teacher ratings provide independent measures of students' academic and social competence. Self-report measures of cigarette, alcohol, and marijuana use are obtained in the questionnaire, enhanced with biochemical measures of CO and EtOH through expired-air samples. Structural modeling techniques are used to test hypotheses about relations between competence and reduce risk for substance use. The results will have implications for theoretical models of adolescent substance use and the design of substance abuse prevention programs.